AI tools turn routine chest scans into screening devices for bone and heart disease
Two artificial intelligence systems presented at RSNA 2025 extract diagnostic information from standard chest imaging without requiring additional scans. One identifies osteoporosis risk from chest X-rays. The other calculates cardiovascular risk from routine chest CT examinations.
An estimated 500 million people worldwide have osteoporosis or its precursor, osteopenia, yet most remain unaware. The standard diagnostic test - dual-energy X-ray absorptiometry (DEXA) - faces barriers including limited availability in underserved regions, long wait times, technologist shortages, and cost.
Bone health screening from chest X-rays
Chest X-rays account for roughly 45% of all radiographic exams performed by physicians. Researchers at the University of Ulsan College of Medicine tested an AI tool called Osteo Signal on 77,677 chest X-rays from 44,773 patients who had undergone DEXA scans within six months. The tool achieved 90% sensitivity and 81% specificity in identifying patients at risk for bone mass reduction.
The AI uses deep learning to identify radiographic patterns associated with osteoporosis risk and generates a numeric score automatically added to radiology records. Osteo Signal has received clearance for commercial use in Korea, Vietnam, and Indonesia. The developer, Promedius, has applied for CE Mark and FDA clearance, with approval expected in late 2026.
Up to 80% of people who suffer a fragility fracture are never diagnosed or treated for the underlying bone disease, according to the International Osteoporosis Foundation. Early identification through routine imaging could reduce that gap.
Cardiovascular risk from non-cardiac CT scans
Researchers in Azerbaijan developed an AI tool that calculates coronary artery calcification (CAC) scores from routine chest CT examinations - scans ordered for other reasons that may contain visible coronary calcification.
A retrospective study of 8,547 routine chest CT scans from six academic centers found that 2,847 patients with previously unclassified risk were reclassified to higher categories. Of the total, 24.3% had their risk reclassified and 18.2% received interventions.
The tool achieved 94.8% sensitivity and 97.2% specificity across all severity categories. The analysis takes 12 seconds to complete. The developers emphasize the tool is not a replacement for dedicated cardiac CT but identifies at-risk asymptomatic patients who might otherwise be overlooked.
Early identification enables timely initiation of preventive therapies without additional radiation exposure or procedural burden. The tool also optimizes resources by providing value from existing imaging infrastructure without significant workflow changes.
The AI quantification tool is currently in the research and validation stage. Developers plan to partner with established medical imaging or AI companies for commercialization.
Healthcare professionals interested in AI for Healthcare applications should understand how these systems extract diagnostic value from data already collected during routine clinical practice.
Your membership also unlocks:
